[4/5] net/nfp: use flag bits to control parsing meta data
Checks
Commit Message
From: Long Wu <long.wu@corigine.com>
Use flag bits to indicate whether meta data exists and determine
whether it needs to be parsed.
Signed-off-by: Long Wu <long.wu@corigine.com>
Reviewed-by: Chaoyong He <chaoyong.he@corigine.com>
---
drivers/net/nfp/nfp_net_meta.c | 22 ++++++++++++++++++----
drivers/net/nfp/nfp_net_meta.h | 1 -
drivers/net/nfp/nfp_rxtx.c | 2 +-
3 files changed, 19 insertions(+), 6 deletions(-)
@@ -27,13 +27,16 @@ nfp_net_meta_parse_chained(uint8_t *meta_base,
meta_info = rte_be_to_cpu_32(meta_header);
meta_offset = meta_base + 4;
+ meta->flags = 0;
for (; meta_info != 0; meta_info >>= NFP_NET_META_FIELD_SIZE, meta_offset += 4) {
switch (meta_info & NFP_NET_META_FIELD_MASK) {
case NFP_NET_META_PORTID:
+ meta->flags |= (1 << NFP_NET_META_PORTID);
meta->port_id = rte_be_to_cpu_32(*(rte_be32_t *)meta_offset);
break;
case NFP_NET_META_HASH:
+ meta->flags |= (1 << NFP_NET_META_HASH);
/* Next field type is about the hash type */
meta_info >>= NFP_NET_META_FIELD_SIZE;
/* Hash value is in the data field */
@@ -41,6 +44,7 @@ nfp_net_meta_parse_chained(uint8_t *meta_base,
meta->hash_type = meta_info & NFP_NET_META_FIELD_MASK;
break;
case NFP_NET_META_VLAN:
+ meta->flags |= (1 << NFP_NET_META_VLAN);
vlan_info = rte_be_to_cpu_32(*(rte_be32_t *)meta_offset);
meta->vlan[meta->vlan_layer].offload =
vlan_info >> NFP_NET_META_VLAN_OFFLOAD;
@@ -50,8 +54,8 @@ nfp_net_meta_parse_chained(uint8_t *meta_base,
meta->vlan_layer++;
break;
case NFP_NET_META_IPSEC:
+ meta->flags |= (1 << NFP_NET_META_IPSEC);
meta->sa_idx = rte_be_to_cpu_32(*(rte_be32_t *)meta_offset);
- meta->ipsec_type = meta_info & NFP_NET_META_FIELD_MASK;
break;
case NFP_NET_META_MARK:
meta->flags |= (1 << NFP_NET_META_MARK);
@@ -77,6 +81,7 @@ nfp_net_meta_parse_single(uint8_t *meta_base,
rte_be32_t meta_header,
struct nfp_net_meta_parsed *meta)
{
+ meta->flags |= (1 << NFP_NET_META_HASH);
meta->hash_type = rte_be_to_cpu_32(meta_header);
meta->hash = rte_be_to_cpu_32(*(rte_be32_t *)(meta_base + 4));
}
@@ -92,6 +97,9 @@ nfp_net_meta_parse_hash(const struct nfp_net_meta_parsed *meta,
if ((hw->super.ctrl & NFP_NET_CFG_CTRL_RSS_ANY) == 0)
return;
+ if (((meta->flags >> NFP_NET_META_HASH) & 0x1) == 0)
+ return;
+
mbuf->hash.rss = meta->hash;
mbuf->ol_flags |= RTE_MBUF_F_RX_RSS_HASH;
}
@@ -109,6 +117,9 @@ nfp_net_meta_parse_vlan(const struct nfp_net_meta_parsed *meta,
if ((ctrl & (NFP_NET_CFG_CTRL_RXVLAN | NFP_NET_CFG_CTRL_RXVLAN_V2)) == 0)
return;
+ if (((meta->flags >> NFP_NET_META_VLAN) & 0x1) == 0)
+ return;
+
/*
* The firmware support two ways to send the VLAN info (with priority) :
* 1. Using the metadata when NFP_NET_CFG_CTRL_RXVLAN_V2 is set,
@@ -156,6 +167,9 @@ nfp_net_meta_parse_qinq(const struct nfp_net_meta_parsed *meta,
(hw->cap & NFP_NET_CFG_CTRL_RXQINQ) == 0)
return;
+ if (((meta->flags >> NFP_NET_META_VLAN) & 0x1) == 0)
+ return;
+
if (meta->vlan_layer < NFP_NET_META_MAX_VLANS)
return;
@@ -184,12 +198,12 @@ nfp_net_meta_parse_ipsec(struct nfp_net_meta_parsed *meta,
struct nfp_net_hw *hw;
struct nfp_tx_ipsec_desc_msg *desc_md;
+ if (((meta->flags >> NFP_NET_META_IPSEC) & 0x1) == 0)
+ return;
+
hw = rxq->hw;
sa_idx = meta->sa_idx;
- if (meta->ipsec_type != NFP_NET_META_IPSEC)
- return;
-
if (sa_idx >= NFP_NET_IPSEC_MAX_SA_CNT) {
mbuf->ol_flags |= RTE_MBUF_F_RX_SEC_OFFLOAD_FAILED;
} else {
@@ -77,7 +77,6 @@ struct nfp_net_meta_parsed {
uint32_t mark_id; /**< Mark id value */
uint16_t flags; /**< Bitmap to indicate if meta exist */
uint8_t hash_type; /**< RSS hash type */
- uint8_t ipsec_type; /**< IPsec type */
uint8_t vlan_layer; /**< The valid number of value in @vlan[] */
/**
* Holds information parses from NFP_NET_META_VLAN.
@@ -497,7 +497,7 @@ nfp_net_recv_pkts(void *rx_queue,
mb->next = NULL;
mb->port = rxq->port_id;
- struct nfp_net_meta_parsed meta = {};
+ struct nfp_net_meta_parsed meta;
nfp_net_meta_parse(rxds, rxq, hw, mb, &meta);
nfp_net_parse_ptype(rxq, rxds, mb);